Ruth Kellerhals (Université de Fribourg)
Cusp density and (in-)commensurability of hyperbolic 3-orbifolds abstract
Abstract:
We consider certain infinite families of 1-cusped hyperbolic 3-orbifolds and show that their elements are pairwise incommensurable. An important feature is their non-arithmeticity. In this way, their cusp density serves as a commensurability invariant and will play an particular role.
10:20 • Université de Fribourg, Chemin du Musée 23, Lecture room 2.52 Physics
Dr. Irene Pasquinelli (Institut de Mathématiques de Jussieu)
Deligne-Mostow lattices: fundamental domains and line arrangements abstract
Abstract:
We will talk about lattices in the group PU(n,1) of holomorphic isometries of complex hyperbolic space. A well known class is that of the Deligne-Mostow lattices. I will describe different interpretations of these lattices and explain how one can use Thurston\'s interpretation to build fundamental domains for them. Then I will explain how Hirzebruch\'s line arrangements construction of the Deligne-Mostow lattices is related to the fundamental domains. Finally I will tell you how we used this connection to go a step forward towards giving a complex equivalent to the hybrid construction. Part of these results are in a joint work with Elisha Falbel.
10:30 • UZH Irchel, Winterthurerstrasse 190, Zürich, Building Y27, Room H 28
Kaspar Rufibach (Roche)
Title T.B.A.
11:00 • Universität Basel, Spiegelgasse 1, Room 00.003
Maria Teresa Chiri (Penn State University)
14:15 • Universität Basel, Zoom Seminar
Prof. Dr. Amin Shokrollahi (EPFL)
Chord Signaling abstract
Abstract:<button>Video<i class="fa fa-play-circle"></button> (**This eSeminar will also be live-streamed on Zoom, using the same meeting details as previous seminars. If you do not have meeting details, please contact karan.khathuria@math.uzh.ch **) <BR> <BR> Communication of data on electrical wires between chips is fast gaining prominence in the electronics industry. Because most of the components of the transmitter and the receiver of such links are analog, rather than digital, they don\'t benefit as much from Moore\'s law. On the other hand, the need to transmit data ever faster calls for higher rates of transmission over existing electrical wires. Since in this type of communication, noise is highly frequency dependent, higher transmission rates lead to much higher noise, and therefore a much higher growth of power consumption than linear. The industry has long recognised this problem as the "Interconnect bottleneck". Fundamental solutions to this important problem have remained elusive, however. <BR> A look at the capacity of these channels reveals that today we are only transmitting at anywhere between 1% to 4% of the capacity. Therefore, at least on the surface, there is a lot to be gained by applying methods from communication theory to this problem. However, unlike many other systems such as wireless, DSL, satellite, or optical communication, the constraints on the chip-to-chip communication system are very different: transmission rates are typically 1000 times those encountered in wireless communication. On the other hand, the energy consumed for the transmission and recovery of each bit is about 1000 times less than what is customary in wireless. Also, latency requirements are extremely stringent, allowing only latencies up to very few nanoseconds. Therefore, it is not possible to use fancy processing methods. <BR> In this talk I will introduce a new modulation scheme for chip-to-chip communication which we call chordal codes. These codes are somewhat reminiscent of spatial MIMO systems, and provide a first step towards a better utilisation of the available communication bandwidth between chips. Current implementations of systems based on these codes show a large reduction of total power of the communication PHY and a large increase of the communication speed compared to other classical systems.
15:00 • UZH Irchel, Winterthurerstrasse 190, Zürich, Building Y27, Room H 28
Dr. Longting Wu (ETH Zürich)
A recursive formula for Gromov-Witten invariants of local Fano threefolds abstract
Abstract:
In this talk, we will consider the Gromov-Witten invariants of three Fano threefolds O_{P^2}(-1), O_{P^2}(-2) and O_{P^1\\times P^1}(-1,-1). We will present a simple uniform recursive formula which completely determine their (primary) Gromov-Witten invariants for all genera. This is a work in progress with Pierrick Bousseau.
15:00 • Zoom
Francesca Carocci (Imperial College London)
Log and Tropical geometry: an application desingularization of low genus stable maps to projective space abstract
Abstract:
In the first part of the talk I will recall a few basic notions of toric geometry and try to explain how log geometry generalises them. Afterwords, we will see how this framework is helpful to interpret blow ups in a modular fashion. Finally, we will see in the examples of maps to projective space of low genus how these techniques apply to give a modular resolution of the main component of Kontsevich\'s space of stable maps.
15:15 • EPF Lausanne, MED 2 1522
Prof. Dr. Jakob Zech (Uni Heidelberg, Germany)
Sparse approximation of triangular transports on bounded domains abstract
Abstract:
Transport maps coupling two different measures can be used to sample from arbitrarily complex distributions. One of the main applicationsof this approach concerns Bayesian inference, where sampling from a posterior distribution facilitates making predictions based on partialand noisy measurments. In this talk we investigate the approximation of triangular transports $T:[-1,1]^d\\to [-1,1]^d$ on the$d$-dimensional unit cube by polynomial expansions and ReLU networks. Specifically, given a reference and a target probability measure with positive and analytic Lebesgue densities on $[-1,1]^d$,we show that the unique Knothe-Rosenblatt transport, which pushes forward the reference to the target, can be approximated at an exponential rate in case $d<\\infty$. These results are generalized to $d=\\infty$, within a setting which incorporates many posterior densities occurring in PDE-driven Bayesian inverse problems. In theinfinite dimensional case ($d=\\infty$) we verify an algebraic convergence rate, which shows that the curse of dimensionality can be overcome.
Dr. Claire Burrin (ETH Zürich, Switzerland)
A sparse equidistribution problem for expanding horocycles on the modular surface abstract
Abstract:
The orbits of the horocycle flow on surfaces are classified: each orbit is either dense or a closed horocycle around a cusp. In particular, expanding closed horocycles are themselves asymptotically dense, and in fact become equidistributed on the surface. The precise rate of equidistribution is of interest; on the modular surface, Zagier observed that a particular rate is equivalent to the Riemann hypothesis being true. In this talk, we explore the asymptotic behavior of evenly spaced points along an expanding closed horocycle on the modular surface. In this problem, the number of sparse points is made to depend on the expansion rate, and the difficulty is that these points are no more invariant under the horocycle flow. This is based on joint work with Uri Shapira and Shucheng Yu.
16:15 • ETH Zentrum, Rämistrasse 101, Zürich, Building HG, Room G 19.1
Emmanuel Abbe (EPF Lausanne)
Lower-bounds for deep learning abstract
Abstract:
I will start by discussing the deep learning framework and the new optimization mechanisms that it puts forward. I will then focus on obtaining lower-bounds for descent algorithms and pose some open problems.
16:15 • EPF Lausanne, Salle CO1
Prof. Dr. Elizaveta (Liza) Levina (University of Michigan, Ann Arbor, USA)
Hierarchical community detection by recursive partitioning abstract
Abstract:
Community detection in networks has been extensively studied in the form of finding a single partition into a “correct” number of communities. In large networks, however, a multi-scale hierarchy of communities is much more realistic. We show that a hierarchical tree of communities, obviously more interpretable, is also potentially more accurate and more computationally efficient. We construct this tree with a simple top-down recursive algorithm, at each step splitting the nodes into two communities with a non-iterative spectral algorithm, until a stopping rule suggests there are no more communities. The algorithm is model-free, extremely fast, and requires no tuning other than selecting a stopping rule. We propose a natural model for this setting, a binary tree stochastic block model, and prove that the algorithm correctly recovers the entire community tree under relatively mild assumptions. As a by-product, we obtain explicit and intuitive results for fitting the stochastic block model under model misspecification. We illustrate the algorithm on a statistics papers dataset constructing a highly interpretable tree of statistics research communities, and on a network based on gene co-occurrence in research papers on anemia.Joint work with Tianxi Li, Lihua Lei, Sharmodeep Bhattacharyya, Koen van de Berge, Purnamrita Sarkar, and Peter Bickel.